Exploration of the spatial differentiation of the retail industry based on large-scale geospatial data is of great significance for urban development. In the recent years, POI data has become an important data source for studying urban dynamics. POI data abstracts retail stores as a point on the map, and the data are of wide coverage and high fineness. These advantages make the POI data an ideal dataset for micro- analysis of urban retail commercial structure and their spatial distribution mechanism. Based on the data of 47 026 retail outlets in Guangzhou, we explored the driving mechanism of the spatial differentiation of the retail formats. By building an indicator system, we investigated the factors potentially affecting the spatial differentiation of the retail industry, which include population density, business conditions, public transportation convenience, format richness, and rent. Based on information entropy, kernel density function, and spatial regression, we analyzed the main influencing factors of the retail differentiation. Further, we divided the retail outlets by different urban areas and different retail formats, and conducted spatial regression analysis based on the same influencing factors to compare the main driving factors of different retail formats. Results shows that: (1) Demand, location, competition, and cost were the main driving force of the spatial differentiation of retail industry. Furthermore, because of the heterogeneity of the retail formats and the spatial heterogeneity of the city, there was also heterogeneity in the driving mechanism of the spatial differentiation in the retail industry. (2) There were significant spatial differences of the influencing factors. The inner circle of the city had higher population density, better accessibility, better business conditions, and higher format richness, and also higher land rent. There was a significant spatial differentiation between the old city area and the suburbs. (3) Compared with the traditional OLS regression method, the spatial regression method revealed the spatial distribution mechanism of the retail industry more accurately. The spatial error model revealed significant heterogeneity in the factors that influence the spatial agglomeration of the retail industry. Population density was the core driving force of retail spatial differentiation. Public transport convenience, business conditions, and format richness also had a positive effect on retail agglomeration, while the impact of rent was weak. The main driving factors of different retail formats and outlets located in different urban circle were different. Population density was the core factor, while the influence of other factors showed significant differences.